机器学习解决简答题自动评分问题的发展综述OA

Automatic Short Answer Grading(ASAG)is an important research direction in smart education,which focuses on how to extract features for comparison and grading from reference answers,grading criteria and student response information,and to obtain reasonable grading for student response by building models and optimizing assessment metrics.Natural Language Processing(NLP)is mostly used in the stages of data pre-processing and model building,and Machine Learning has emerged as a mainstream in recent years.It comprehensively summarizes the research and development for ASAG.Firstly,five solutions of ASAG are sorted out and summarized,a focused summary of Machine Learning based on ASAG solutions is presented,the differences between Chinese and English implementations of ASAG is analyzed,and the concerns of each solution and relevant mainstream algorithms are compared and summarized.Then,it compares the main algorithm features of ASAG and their effectiveness on typical datasets.Finally,the current problems and challenges faced by the research on ASAG and the future trends are described.